[Bug] Unexpected high token usage spike on single prompt request

Question

**Bug Description**
I had one prompt to read some docs (and a screenshot) to generate a design doc. That prompt used my 5-hour usage quote from 0% to 25%. 
A second prompt is provide some feedback to review the design doc, the 5-hour usage limit jumped from 25% to 60%.

I am on the Pro plan.

This situation seems to be getting worse. This is not the experience I had couple of weeks ago. I had some very decent usage 2 weeks ago. This is a completely different. Please help!

**Environment Info**